Open-source document reader
OpenReader turns EPUB, PDF, TXT, Markdown, and DOCX into a synchronized read-along surface, reading your original file in a native viewer with multilingual text-to-speech, language-aware highlighting, and audiobook export. It’s open source, and entirely yours to self-host.
The flow
Drop an EPUB, PDF, TXT, Markdown, or DOCX into your library, or import one straight from the server, and it stays organized for every session after.
Choose a provider and model: hosted OpenAI, Replicate, or DeepInfra, or your own self-hosted Kokoro, KittenTTS, or Orpheus server. Set the document language, choose a compatible voice, and adjust the speed to your pace.
Follow word-level highlighting right on the original page as it plays, pick up where you left off on any device, and export a chaptered m4b or mp3 audiobook for the road.
Under the hood
Your file renders in a built-in EPUB and PDF reader, never flattened to plain text. Layout-aware parsing (PP-DocLayoutV3, ONNX) maps the structure underneath, so read-along highlighting follows the true reading order, even in dense, multi-column PDFs.
ONNX Whisper alignment through a JetStream-backed compute worker maps each spoken word back to the page, so the cursor tracks speech precisely.
Choose a document language for language-aware narration, highlighting, and compatible voice selection.
Render chaptered m4b and mp3 files with resumable processing, ready for any offline player you already use.
Reading position, queue state, and per-document settings sync across browser sessions and devices through your account.
Run on embedded SeaweedFS or any S3-compatible bucket, back it with SQLite or Postgres, and ship it with Docker on amd64 or arm64, with built-in auth and automatic startup migrations.
Open source · MIT
Deploy OpenReader for yourself or your team in minutes. The docs cover Docker, provider integration, object storage, and the external compute worker. Every piece is yours to host.
